---
name: linkedin-export-profile-writer
description: Use when a user wants to turn a LinkedIn data export into polished Markdown assets such as an AI profile, speaker sheet, executive bio, or a use-cases guide. This skill is for extracting signal from LinkedIn export CSV and HTML files, creating privacy-aware narrative documents, and rewriting them so they sound authored rather than AI-generated.
---

# LinkedIn Export Profile Writer

Use this skill when the input is a LinkedIn data export and the user wants finished written assets rather than raw analysis. The job is to inspect the export, identify the highest-signal files, and turn them into clean Markdown documents that are portable, privacy-aware, and written in a human narrative style.

## What This Skill Produces
The most common outputs are:

- an AI-ready personal profile
- a speaker sheet
- an executive or recruiter bio
- a playbook of other useful things to do with the export
- a redaction or privacy-oriented summary

These should remain Markdown unless the user asks for another format.

## Source Files To Prioritize
Start with the files that define the person professionally:

- `Profile.csv`
- `Positions.csv`
- `Education.csv`
- `Skills.csv`
- `Certifications.csv`
- `Honors.csv`
- `Organizations.csv`
- `Volunteering.csv`
- `Courses.csv`
- `Causes You Care About.csv`

Use these to deepen or extend the work when relevant:

- `Shares.csv`
- `Comments.csv`
- `Articles/Articles/*.html`
- `Learning.csv`
- `Events.csv`
- `Company Follows.csv`
- `Hashtag_Follows.csv`
- `Endorsement_Received_Info.csv`

Treat these as sensitive by default and do not quote or summarize them directly into public-facing outputs unless the user clearly asks for it:

- `Connections.csv`
- `Email Addresses.csv`
- `PhoneNumbers.csv`
- `Jobs/*.csv`
- `Job Applicant Saved Screening Question Responses.csv`
- `Verifications/Verifications.csv`
- `Inferences_about_you.csv`
- `Ad_Targeting.csv`

## Workflow
First inspect the export structure with `rg --files`, then open the high-signal CSVs to understand both field names and actual content. Do not ask the user questions that can be answered by reading the files. Before writing anything, decide which files belong in the main narrative and which should stay out for privacy or quality reasons.

When creating the first draft, write for signal, not coverage. Use current role, career progression, education, credentials, public-facing themes, and service history to establish the through-line. The strongest outputs usually center on role progression, leadership scope, technical domain, and recurring themes visible across profile text and activity.

After drafting, do a second pass specifically for writing quality. This is not optional. Most first drafts from structured data read too list-heavy and too synthetic. Rewrite them into authored prose.

## Writing Standard
The writing should sound like a person wrote it after reading the data, not like a model expanding bullet points. Prefer paragraphs over lists. Use bullets only when the format genuinely benefits from scanning, such as sample session titles or short enumerations where list structure is the clearest form.

Avoid these patterns:

- excessive section stacking with thin content
- headings followed immediately by bullets with no narrative
- repetitive labels such as "What it supports" or "Concrete outputs" repeated over and over
- generic phrases like "leverages AI to drive outcomes" or "brings a unique perspective"
- inflated, vague, or recruiter-speak language

Prefer these patterns:

- strong opening paragraph that frames the person or document
- sections that each have a clear job and contain real prose
- transitions that explain why a section matters
- specific but restrained claims grounded in the export
- a tone that is direct, credible, and readable

## Recommended Output Shapes
For an AI profile, combine a short third-person framing with a first-person section that tells an AI assistant how to help. Include current work, career arc, areas of strength, education and credentials, service or community involvement, and a short note on intended use.

For a speaker sheet, include a strong overview, short/medium/long bio options, topic framing, event fit, credibility, a host introduction, and booking notes. A short list of sample talk titles is acceptable here because it is genuinely useful.

For a use-cases guide, do not produce a giant inventory by file. Group the material into a few meaningful narratives such as AI context, career materials, content mining, network analysis, knowledge base creation, and privacy review. Explain why each category matters and which source files support it.

## Privacy Rules
Assume the user may want broad professional detail without wanting raw sensitive fields copied into the final documents. By default:

- include current and past roles, education, credentials, honors, community work, and public themes
- exclude birth date, phone numbers, email dumps, instant messenger handles, verification data, application answers, inferred targeting traits, and connection-level details
- if contact information is needed, prefer public-facing website or public profile links already present in the export

If the user explicitly asks for broader inclusion, still distinguish between useful narrative detail and raw data that should not be pasted verbatim.

## Quality Check Before Finishing
Before returning the files, read them back and check for these issues:

- Does the document sound outline-driven or actually written?
- Are there too many bullets relative to paragraphs?
- Are claims grounded in the export?
- Did any sensitive raw fields slip in?
- Does each section earn its place?

If the answer to any of those is no, rewrite.